pith. sign in

arxiv: 2511.02659 · v3 · submitted 2025-11-04 · 💻 cs.LG · cs.AI· cs.CE· cs.NA· math.NA

In Situ Training of Implicit Neural Compressors for Scientific Simulations via Sketch-Based Regularization

Cooper Simpson , Stephen Becker , Alireza Doostan This is my paper

Pith reviewed 2026-05-18 01:11 UTC · model grok-4.3

classification 💻 cs.LG cs.AIcs.CEcs.NAmath.NA
keywords implicit neural representationsin situ trainingsketchingcatastrophic forgettingscientific data compressionhypernetworkscontinual learning
0
0 comments X

The pith

Sketching lets in-situ neural compressors approximately match offline training performance.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops an in-situ training approach for implicit neural representations used to compress scientific simulation data. Instead of storing all data and training offline, it trains the model continuously as the simulation runs using a small memory buffer. The buffer holds some full data points and uses sketched versions of others as a regularizer to avoid the model forgetting what it learned earlier. Theory based on the Johnson-Lindenstrauss lemma supports why sketching works for this. Experiments on two- and three-dimensional simulations show that this method achieves reconstruction quality close to full offline training at high compression ratios.

Core claim

The authors show that a hypernetwork-based implicit neural compressor can be trained in situ on streaming simulation data by regularizing with sketched samples in a limited memory buffer, yielding reconstruction performance that approximately matches the equivalent offline training method.

What carries the argument

The sketch-based regularizer applied to a mixed buffer of full and sketched samples during continual training of the hypernetwork.

Load-bearing premise

That the sketched samples in the buffer provide sufficient regularization to prevent catastrophic forgetting in the continual in-situ training.

What would settle it

A direct comparison where the in-situ method with sketching produces markedly higher reconstruction errors or poorer visual quality than offline training on identical simulation data would disprove the main claim.

Figures

Figures reproduced from arXiv: 2511.02659 by Alireza Doostan, Cooper Simpson, Stephen Becker.

Figure 1
Figure 1. Figure 1: The compression approach of this study applies to all mesh/geometry types, including (uniform) structured (left), [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Compression results for INRs trained offline. See Section 3 for details on the datasets. [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Our in situ INC training approach relies on a sketched data buffer to avoid catastrophic forgetting. 6 [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: INC network structure. by the initialization of the target INR, and the weights of the last linear layer are scaled by a relatively small factor. We should note that many more tricks and techniques exist in the literature for both hypernetworks and implicit neural representation, much of which was discussed in Section 1.1. Our architecture may benefit from these alterations. As an example, a more sophistic… view at source ↗
Figure 5
Figure 5. Figure 5: Reconstruction comparison of the Ignition data at snapshot [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Reconstruction comparison of the Channel dataset on all three channel flow velocities ( [PITH_FULL_IMAGE:figures/full_fig_p015_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Training and testing error on the Ignition data for the InSitu-FJLT case. Training error is taken at the end of the [PITH_FULL_IMAGE:figures/full_fig_p016_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: In situ Sketch sample factor performance comparison at fixed compression rate on Neuron dataset. On the theoretical side, much more can be done to explain the empirical success of small sketches in preventing catastrophic forgetting. The relevant question to ask is how large a sketch needs to be to serve as an effective regularizer? Rigorously answering this question would lead to a more precise setting of… view at source ↗
read the original abstract

Focusing on implicit neural representations, we present a novel in situ training protocol that employs limited memory buffers of full and sketched data samples, where the sketched data are leveraged to prevent catastrophic forgetting. The theoretical motivation for our use of sketching as a regularizer is presented via a simple Johnson-Lindenstrauss-informed result. While our methods may be of wider interest in the field of continual learning, we specifically target in situ neural compression using implicit neural representation-based hypernetworks. We evaluate our method on a variety of complex simulation data in two and three dimensions, over long time horizons, and across unstructured grids and non-Cartesian geometries. On these tasks, we show strong reconstruction performance at high compression rates. Most importantly, we demonstrate that sketching enables the presented in situ scheme to approximately match the performance of the equivalent offline method.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces an in-situ training protocol for implicit neural representation (INR) compressors based on hypernetworks. It maintains a limited-memory buffer containing both full-resolution and sketched samples from simulation trajectories; the sketched samples are used as a regularizer to mitigate catastrophic forgetting during continual updates. A Johnson-Lindenstrauss lemma is invoked to motivate that distance preservation in the sketched buffer approximately preserves the offline training objective. Experiments on 2-D and 3-D scientific simulation data (including unstructured grids and non-Cartesian geometries) over long time horizons report reconstruction quality at high compression ratios that approximately matches the corresponding offline baseline.

Significance. If the empirical match to offline performance is shown to be attributable to the sketching mechanism rather than buffer size or schedule choices, the approach would provide a practical route to memory-efficient, high-fidelity compression of large-scale simulation data that cannot be stored offline. The combination of hypernetwork-generated INRs with JL-motivated sketching for continual learning is a concrete contribution that could influence both scientific data management and broader continual-learning research.

major comments (2)
  1. [theoretical motivation] Theoretical motivation section: the JL lemma is stated to bound Euclidean distances between full and sketched samples in data space, yet the training objective is a reconstruction loss on an INR whose parameters are produced by a hypernetwork. No derivation is given showing that distance preservation in the input samples implies bounded drift in the hypernetwork weights or in the decoded field values, particularly under non-uniform sampling measures on unstructured grids. This gap is load-bearing for the central claim that sketching, rather than other factors, enables the in-situ scheme to match offline performance.
  2. [evaluation sections] Experimental protocol and results: the abstract and evaluation sections report strong reconstruction performance but provide neither quantitative error bars across multiple runs, nor ablations that isolate the contribution of the sketched buffer versus buffer size or learning-rate schedule. Without these controls it is difficult to attribute the observed match to offline performance specifically to the JL-informed regularizer.
minor comments (2)
  1. [method] Notation for the hypernetwork and INR decoder should be introduced with explicit variable definitions before the loss function is written.
  2. [figures] Figure captions for the reconstruction visualizations should state the compression ratio and the norm used for the reported error.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which help clarify the contributions and limitations of our work. We address each major comment below.

read point-by-point responses

  1. Referee: [theoretical motivation] Theoretical motivation section: the JL lemma is stated to bound Euclidean distances between full and sketched samples in data space, yet the training objective is a reconstruction loss on an INR whose parameters are produced by a hypernetwork. No derivation is given showing that distance preservation in the input samples implies bounded drift in the hypernetwork weights or in the decoded field values, particularly under non-uniform sampling measures on unstructured grids. This gap is load-bearing for the central claim that sketching, rather than other factors, enables the in-situ scheme to match offline performance.

    Authors: We appreciate this observation regarding the theoretical motivation. Our use of the Johnson-Lindenstrauss lemma is intended as a motivational tool to indicate that sketching preserves important geometric properties of the data, thereby helping to approximate the offline training objective in the in-situ setting. We acknowledge that we do not provide a full derivation connecting data-space distance preservation to bounds on hypernetwork weight drift or decoded field values under non-uniform sampling. In the revision, we will add a more explicit discussion of this connection, including a heuristic argument, and note that a rigorous proof remains an open direction for future work. revision: partial

  2. Referee: [evaluation sections] Experimental protocol and results: the abstract and evaluation sections report strong reconstruction performance but provide neither quantitative error bars across multiple runs, nor ablations that isolate the contribution of the sketched buffer versus buffer size or learning-rate schedule. Without these controls it is difficult to attribute the observed match to offline performance specifically to the JL-informed regularizer.

    Authors: We agree that the current experimental presentation lacks sufficient controls to fully isolate the effect of the sketched regularizer. To address this, we will revise the evaluation sections to include error bars computed over multiple random seeds and introduce ablation experiments that compare variants with and without sketching while keeping buffer size and optimization schedules fixed. This will help attribute performance differences more clearly to the sketching mechanism. revision: yes

Circularity Check

0 steps flagged

No significant circularity; central claims rest on external JL lemma and independent empirical evaluations

full rationale

The paper's theoretical motivation invokes the Johnson-Lindenstrauss lemma as an external mathematical fact to justify sketching as a regularizer against catastrophic forgetting. Performance claims that sketching enables the in-situ scheme to approximately match offline results are supported by evaluations on independent simulation datasets across 2D/3D, long horizons, unstructured grids, and non-Cartesian geometries rather than any reduction to fitted parameters defined inside the paper or self-citation chains. No load-bearing derivation step reduces by construction to its own inputs, self-definition, or renamed known results.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The approach rests on the applicability of sketching as a regularizer and the sufficiency of limited buffers; no new physical entities are introduced.

free parameters (1)
  • sketch dimension / buffer size
    Size of the sketched data buffer is a tunable hyperparameter that controls the regularization strength and memory footprint.
axioms (1)
  • domain assumption Johnson-Lindenstrauss lemma provides distance preservation sufficient to act as a regularizer against catastrophic forgetting in this continual-learning setting
    Invoked as the theoretical motivation for using sketches in the in-situ protocol.

pith-pipeline@v0.9.0 · 5684 in / 1245 out tokens · 37722 ms · 2026-05-18T01:11:54.376535+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

58 extracted references · 58 canonical work pages

  1. [1]

    Deepsdf: Learning continuous signed distance functions for shape representation

    Jeong Joon Park, Peter Florence, Julian Straub, Richard Newcombe, and Steven Lovegrove. Deepsdf: Learning continuous signed distance functions for shape representation. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 165–174, 2019. 1

    work page 2019

  2. [2]

    Learning shape templates with structured implicit functions

    Kyle Genova, Forrester Cole, Daniel Vlasic, Aaron Sarna, William T Freeman, and Thomas Funkhouser. Learning shape templates with structured implicit functions. InProceedings of the IEEE/CVF international conference on computer vision, pages 7154–7164, 2019. 1

    work page 2019

  3. [3]

    Implicit neural representations with periodic activation functions.Advances in neural information processing systems, 33:7462–7473, 2020

    Vincent Sitzmann, Julien Martel, Alexander Bergman, David Lindell, and Gordon Wetzstein. Implicit neural representations with periodic activation functions.Advances in neural information processing systems, 33:7462–7473, 2020. 1, 2, 4, 11

    work page 2020

  4. [4]

    Implicit neural video compression

    Yunfan Zhang, Ties Van Rozendaal, Johann Brehmer, Markus Nagel, and Taco Cohen. Implicit neural video compression. InICLR Workshop on Deep Generative Models for Highly Structured Data, 2022. 2, 5

    work page 2022

  5. [5]

    Neural implicit flow: a mesh-agnostic dimen- sionality reduction paradigm of spatio-temporal data.Journal of Machine Learning Research, 24(41): 1–60, 2023

    Shaowu Pan, Steven L Brunton, and J Nathan Kutz. Neural implicit flow: a mesh-agnostic dimen- sionality reduction paradigm of spatio-temporal data.Journal of Machine Learning Research, 24(41): 1–60, 2023. 2, 4, 11

    work page 2023

  6. [6]

    Kd-inr: Time-varying volumetric data compression via knowledge distillation-based implicit neural representation.IEEE Transactions on Visualization and Computer Graphics, 2023

    Jun Han, Hao Zheng, and Chongke Bi. Kd-inr: Time-varying volumetric data compression via knowledge distillation-based implicit neural representation.IEEE Transactions on Visualization and Computer Graphics, 2023. 2, 5, 11

    work page 2023

  7. [7]

    Implicit neural compression for aerospace simulation visualisation

    Robert M Sales and Graham Pullan. Implicit neural compression for aerospace simulation visualisation. InAIAA SCITECH 2024 Forum, page 0164, 2024. 2, 5, 11

    work page 2024

  8. [8]

    Dai, and Quoc V

    David Ha, Andrew M. Dai, and Quoc V. Le. Hypernetworks. InInternational Conference on Learning Representations, 2017. 2, 4

    work page 2017

  9. [9]

    Multitask learning.Machine learning, 28(1):41–75, 1997

    Rich Caruana. Multitask learning.Machine learning, 28(1):41–75, 1997. 2

    work page 1997

  10. [10]

    Karniadakis

    Maziar Raissi, Paris Perdikaris, and George E. Karniadakis. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations.Journal of Computational physics, 378:686–707, 2019. 3

    work page 2019

  11. [11]

    Wire: Wavelet implicit neural representations

    Vishwanath Saragadam, Daniel LeJeune, Jasper Tan, Guha Balakrishnan, Ashok Veeraraghavan, and Richard G Baraniuk. Wire: Wavelet implicit neural representations. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18507–18516, 2023. 4

    work page 2023

  12. [12]

    Implicit neural representations and the algebra of complex wavelets

    T Mitchell Roddenberry, Vishwanath Saragadam, Maarten V de Hoop, and Richard Baraniuk. Implicit neural representations and the algebra of complex wavelets. InThe Twelfth International Conference on Learning Representations, 2024. 4

    work page 2024

  13. [13]

    Where do we stand with implicit neural representa- tions? a technical and performance survey.Transactions on Machine Learning Research, 2025

    Amer Essakine, Yanqi Cheng, Chun-Wun Cheng, Lipei Zhang, Zhongying Deng, Lei Zhu, Carola- Bibiane Schönlieb, and Angelica I Aviles-Rivero. Where do we stand with implicit neural representa- tions? a technical and performance survey.Transactions on Machine Learning Research, 2025. ISSN 2835-8856. 4

    work page 2025

  14. [14]

    A brief review of hypernetworks in deep learning.Artificial Intelligence Review, 57(9):1–29, 2024

    Vinod Kumar Chauhan, Jiandong Zhou, Ping Lu, Soheila Molaei, and David A Clifton. A brief review of hypernetworks in deep learning.Artificial Intelligence Review, 57(9):1–29, 2024. 4

    work page 2024

  15. [15]

    Model reduction of dynamical systems on nonlinear manifolds using deep convolutional autoencoders.Journal of Computational Physics, 404:108973, 2020

    Kookjin Lee and Kevin T Carlberg. Model reduction of dynamical systems on nonlinear manifolds using deep convolutional autoencoders.Journal of Computational Physics, 404:108973, 2020. 4 19

    work page 2020

  16. [16]

    Overcoming catastrophic forgetting in neural networks.Proceedings of the national academy of sciences, 114(13): 3521–3526, 2017

    James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, et al. Overcoming catastrophic forgetting in neural networks.Proceedings of the national academy of sciences, 114(13): 3521–3526, 2017. 4

    work page 2017

  17. [17]

    Continual learning with hypernetworks

    Johannes Von Oswald, Christian Henning, Benjamin F Grewe, and João Sacramento. Continual learning with hypernetworks. InInternational Conference on Learning Representations, 2020. 4

    work page 2020

  18. [18]

    Experience replay for continual learning.Advances in neural information processing systems, 32, 2019

    David Rolnick, Arun Ahuja, Jonathan Schwarz, Timothy Lillicrap, and Gregory Wayne. Experience replay for continual learning.Advances in neural information processing systems, 32, 2019. 4

    work page 2019

  19. [19]

    A comprehensive survey of continual learning: Theory, method and application.IEEE Transactions on Pattern Analysis and Machine Intelligence,

    Liyuan Wang, Xingxing Zhang, Hang Su, and Jun Zhu. A comprehensive survey of continual learning: Theory, method and application.IEEE Transactions on Pattern Analysis and Machine Intelligence,

    work page

  20. [20]

    Fixed-rate compressed floating-point arrays.IEEE Transactions on Visualization and Computer Graphics, 20(12):2674–2683, 2014

    Peter Lindstrom. Fixed-rate compressed floating-point arrays.IEEE Transactions on Visualization and Computer Graphics, 20(12):2674–2683, 2014. doi: 10.1109/TVCG.2014.2346458. 4

    work page doi:10.1109/tvcg.2014.2346458 2014

  21. [21]

    Gok, Jiannan Tian, Junjing Deng, Jon C

    Xin Liang, Kai Zhao, Sheng Di, Sihuan Li, Robert Underwood, Ali M. Gok, Jiannan Tian, Junjing Deng, Jon C. Calhoun, Dingwen Tao, Zizhong Chen, and Franck Cappello. Sz3: A modular framework for composing prediction-based error-bounded lossy compressors.IEEE Transactions on Big Data, 9 (2):485–498, 2023. doi: 10.1109/TBDATA.2022.3201176. 4

    work page doi:10.1109/tbdata.2022.3201176 2023

  22. [22]

    Mesh-float-zip (mfz): Manifold harmonic bases for unstructured spatial data compression.Applied Mathematics for Modern Challenges, 2(4):465–489,

    Kevin Doherty, Stephen Becker, and Alireza Doostan. Mesh-float-zip (mfz): Manifold harmonic bases for unstructured spatial data compression.Applied Mathematics for Modern Challenges, 2(4):465–489,

    work page

  23. [23]

    doi: 10.3934/ammc.2024023. 4

    work page doi:10.3934/ammc.2024023

  24. [24]

    A survey on error-bounded lossy compression for scientific datasets.ACM computing surveys, 57(11):1–38, 2025

    Sheng Di, Jinyang Liu, Kai Zhao, Xin Liang, Robert Underwood, Zhaorui Zhang, Milan Shah, Yafan Huang, Jiajun Huang, Xiaodong Yu, et al. A survey on error-bounded lossy compression for scientific datasets.ACM computing surveys, 57(11):1–38, 2025. 4

    work page 2025

  25. [25]

    The approximation of one matrix by another of lower rank.Psychometrika, 1(3):211–218, 1936

    Carl Eckart and Gale Young. The approximation of one matrix by another of lower rank.Psychometrika, 1(3):211–218, 1936. 5

    work page 1936

  26. [26]

    Finding structure with randomness: Probabilistic algorithms for constructing approximate matrix decompositions.SIAM review, 53(2): 217–288, 2011

    Nathan Halko, Per-Gunnar Martinsson, and Joel A Tropp. Finding structure with randomness: Probabilistic algorithms for constructing approximate matrix decompositions.SIAM review, 53(2): 217–288, 2011. 5

    work page 2011

  27. [27]

    Single-pass pca of large high-dimensional data

    Wenjian Yu, Yu Gu, and Jian Li. Single-pass pca of large high-dimensional data. InProceedings of the 26th International Joint Conference on Artificial Intelligence, pages 3350–3356, 2017. 5

    work page 2017

  28. [28]

    Pass-efficient methods for compression of high-dimensional turbulent flow data.Journal of Computational Physics, 423:109704,

    Alec M Dunton, Lluís Jofre, Gianluca Iaccarino, and Alireza Doostan. Pass-efficient methods for compression of high-dimensional turbulent flow data.Journal of Computational Physics, 423:109704,

    work page

  29. [29]

    Heather Pacella, Alec Dunton, Alireza Doostan, and Gianluca Iaccarino. Task-parallel in situ temporal compression of large-scale computational fluid dynamics data.The International Journal of High Performance Computing Applications, 36(3):388–418, 2022. 5

    work page 2022

  30. [30]

    Angran Li, Stephen Becker, and Alireza Doostan. Online randomized interpolative decomposition with a posteriori error estimator for temporal pde data reduction.Computer Methods in Applied Mechanics and Engineering, 434:117538, 2025. 5

    work page 2025

  31. [31]

    Quadconv: Quadrature-based convolutions with applications to non-uniform pde data compression.Journal of Computational Physics, 498:112636, 2024

    Kevin Doherty, Cooper Simpson, Stephen Becker, and Alireza Doostan. Quadconv: Quadrature-based convolutions with applications to non-uniform pde data compression.Journal of Computational Physics, 498:112636, 2024. 5 20

    work page 2024

  32. [32]

    Autoencoding implicit neural representations for image compression

    Tuan Pham, Yibo Yang, and Stephan Mandt. Autoencoding implicit neural representations for image compression. InICML 2023 Workshop Neural Compression: From Information Theory to Applications,

    work page 2023

  33. [33]

    Tinc: Tree-structured implicit neural compression

    Runzhao Yang. Tinc: Tree-structured implicit neural compression. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18517–18526, 2023. 5

    work page 2023

  34. [34]

    Miner: Multiscale implicit neural representation

    Vishwanath Saragadam, Jasper Tan, Guha Balakrishnan, Richard G Baraniuk, and Ashok Veeraragha- van. Miner: Multiscale implicit neural representation. InEuropean Conference on Computer Vision, pages 318–333. Springer, 2022. 5

    work page 2022

  35. [35]

    Ecnr: efficient compressive neural representation of time-varying volumetric datasets

    Kaiyuan Tang and Chaoli Wang. Ecnr: efficient compressive neural representation of time-varying volumetric datasets. In2024 IEEE 17th Pacific Visualization Conference (PacificVis), pages 72–81. IEEE Computer Society, 2024. 5

    work page 2024

  36. [36]

    Implicit neural representations for image compression

    Yannick Strümpler, Janis Postels, Ren Yang, Luc Van Gool, and Federico Tombari. Implicit neural representations for image compression. InEuropean Conference on Computer Vision, pages 74–91. Springer, 2022. 5

    work page 2022

  37. [37]

    Coin: Compression with implicit neural representations

    Emilien Dupont, Adam Golinski, Milad Alizadeh, Yee Whye Teh, and Arnaud Doucet. Coin: Compression with implicit neural representations. InNeural Compression: From Information Theory to Applications–Workshop@ ICLR 2021, 2021. 5

    work page 2021

  38. [38]

    Nerv: Neural representations for videos.Advances in Neural Information Processing Systems, 34:21557–21568, 2021

    Hao Chen, Bo He, Hanyu Wang, Yixuan Ren, Ser Nam Lim, and Abhinav Shrivastava. Nerv: Neural representations for videos.Advances in Neural Information Processing Systems, 34:21557–21568, 2021. 5

    work page 2021

  39. [39]

    Signal compression via neural implicit representations

    Francesca Pistilli, Diego Valsesia, Giulia Fracastoro, and Enrico Magli. Signal compression via neural implicit representations. InICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 3733–3737. IEEE, 2022. 5

    work page 2022

  40. [40]

    Compressive neural representations of volumetric scalar fields

    Yuzhe Lu, Kairong Jiang, Joshua A Levine, and Matthew Berger. Compressive neural representations of volumetric scalar fields. InComputer Graphics Forum, volume 40, pages 135–146. Wiley Online Library, 2021. 5

    work page 2021

  41. [41]

    In situ compression artifact removal in scientific data using deep transfer learning and experience replay.Machine Learning: Science and Technology, 2(2):025010, 12 2020

    Sandeep Madireddy, Ji Hwan Park, Sunwoo Lee, Prasanna Balaprakash, Shinjae Yoo, Wei-keng Liao, Cory D Hauck, M Paul Laiu, and Richard Archibald. In situ compression artifact removal in scientific data using deep transfer learning and experience replay.Machine Learning: Science and Technology, 2(2):025010, 12 2020. doi: 10.1088/2632-2153/abc326. 5

    work page doi:10.1088/2632-2153/abc326 2020

  42. [42]

    Deterministic matrix sketches for low-rank compression of high-dimensional simulation data.arXiv preprint arXiv:2105.01271, 2021

    Alec Michael Dunton and Alireza Doostan. Deterministic matrix sketches for low-rank compression of high-dimensional simulation data.arXiv preprint arXiv:2105.01271, 2021. 7

    work page arXiv 2021

  43. [43]

    The fast johnson–lindenstrauss transform and approximate nearest neighbors.SIAM Journal on computing, 39(1):302–322, 2009

    Nir Ailon and Bernard Chazelle. The fast johnson–lindenstrauss transform and approximate nearest neighbors.SIAM Journal on computing, 39(1):302–322, 2009. 7

    work page 2009

  44. [44]

    Extensions of Lipschitz mappings into a Hilbert space

    William B Johnson and Joram Lindenstrauss. Extensions of Lipschitz mappings into a Hilbert space. Contemporary mathematics, 26(189-206):1, 1984. 7, 8

    work page 1984

  45. [45]

    Blendenpik: Supercharging lapack’s least-squares solver.SIAM Journal on Scientific Computing, 32(3):1217–1236, 2010

    Haim Avron, Petar Maymounkov, and Sivan Toledo. Blendenpik: Supercharging lapack’s least-squares solver.SIAM Journal on Scientific Computing, 32(3):1217–1236, 2010. 8

    work page 2010

  46. [46]

    Spectral estimation from simulations via sketching.Journal of Computational Physics, 447:110686, 2021

    Zhishen Huang and Stephen Becker. Spectral estimation from simulations via sketching.Journal of Computational Physics, 447:110686, 2021. 8

    work page 2021

  47. [47]

    Springer, New York, 2 edition, 2012

    John Lee.Introduction to Smooth manifolds, 2nd ed. Springer, New York, 2 edition, 2012. 8

    work page 2012

  48. [48]

    Random projections of smooth manifolds.Foundations of computational mathematics, 9(1):51–77, 2009

    Richard G Baraniuk and Michael B Wakin. Random projections of smooth manifolds.Foundations of computational mathematics, 9(1):51–77, 2009. 9 21

    work page 2009

  49. [49]

    Armin Eftekhari and Michael B. Wakin. New analysis of manifold embeddings and signal recovery from compressive measurements.Applied and Computational Harmonic Analysis, 39(1):67–109, 2015. ISSN 1063-5203. doi: https://doi.org/10.1016/j.acha.2014.08.005. 9

    work page doi:10.1016/j.acha.2014.08.005 2015

  50. [50]

    Woodruff

    David P. Woodruff. Sketching as a tool for numerical linear algebra.Foundations and Trends®in Theoretical Computer Science, 10(1–2):1–157, 2014. 10

    work page 2014

  51. [51]

    Modeling intracellular transport and traffic jam in 3D neurons using PDE-constrained optimization.Journal of Mechanics, 38:44–59, 2022

    Angran Li and Yongjie Jessica Zhang. Modeling intracellular transport and traffic jam in 3D neurons using PDE-constrained optimization.Journal of Mechanics, 38:44–59, 2022. 10

    work page 2022

  52. [52]

    A web services accessible database of turbulent channel flow and its use for testing a new integral wall model for les.Journal of Turbulence, 17(2):181–215, 2016

    J Graham, K Kanov, XIA Yang, M Lee, N Malaya, CC Lalescu, R Burns, G Eyink, A Szalay, RD Moser, et al. A web services accessible database of turbulent channel flow and its use for testing a new integral wall model for les.Journal of Turbulence, 17(2):181–215, 2016. 10

    work page 2016

  53. [53]

    A public turbulence database cluster and applications to study lagrangian evolution of velocity increments in turbulence.Journal of Turbulence, 9:N31, 2008

    Yi Li, Eric Perlman, Minping Wan, Yunke Yang, Charles Meneveau, Randal Burns, Shiyi Chen, Alexander Szalay, and Gregory Eyink. A public turbulence database cluster and applications to study lagrangian evolution of velocity increments in turbulence.Journal of Turbulence, 9:N31, 2008. 10

    work page 2008

  54. [54]

    PyTorch 2: Faster Machine Learning Through Dynamic Python Bytecode Transforma- tion and Graph Compilation

    Jason Ansel, Edward Yang, Horace He, Natalia Gimelshein, Animesh Jain, Michael Voznesensky, Bin Bao, Peter Bell, David Berard, Evgeni Burovski, Geeta Chauhan, Anjali Chourdia, Will Constable, Alban Desmaison, Zachary DeVito, Elias Ellison, Will Feng, Jiong Gong, Michael Gschwind, Brian Hirsh, Sherlock Huang, Kshiteej Kalambarkar, Laurent Kirsch, Michael L...

    work page doi:10.1145/3620665.3640366 2024

  55. [55]

    Implicit-Neural-Compression

    Cooper Simpson. Implicit-Neural-Compression. https://github.com/RS-Coop/ Implicit-Neural-Compression, 2024. 11

    work page 2024

  56. [56]

    IEEE Standard for Binary Floating-Point Arithmetic

    IEEE. IEEE Standard for Binary Floating-Point Arithmetic. Technical Report ANSI/IEEE Std 754-1985, IEEE Standards Association, 1985. 11

    work page 1985

  57. [57]

    An algorithm for finding intrinsic dimensionality of data

    Keinosuke Fukunaga and David R Olsen. An algorithm for finding intrinsic dimensionality of data. IEEE Transactions on computers, 100(2):176–183, 1971. 13

    work page 1971

  58. [58]

    scikit-dimension, ver

    Jonathan Bac. scikit-dimension, ver. 0.3.4. https://scikit-dimension.readthedocs.io/, 2021. 13 22

    work page 2021